Professional Documents
Culture Documents
Crux v12n03 Mar
Crux v12n03 Mar
Crux v12n03 Mar
Mathematicorum
Published by the Canadian Mathematical Society.
http://crux.math.ca/
CONTENTS
The Olympiad Corner: 73 M.S. Klamkin 39
Problems: 1121-1130 . . . 50
Solutions: 990-993, 995-998 . . . . . . . 52
A message from the Editor . . . . • • • . • • . 65
- 38 -
- 39 -
M.S. KLAMKIN
I dedicate this Corner to Leo Sauve who has recently retired as editor
after starting this journal 11 years ago. I will miss his editing of this
corner greatly. For those colleagues who wish to drop him a note, his address
is 2206 E, Halifax Drive, Ottawa, Ontario, K1G 2W6.
I give one new problem set this month: the Third Round of the 1982
Leningrad High School Olympiad (my thanks to Alex Merkujev for its transmittal
and to Larry Glasser for its translation). As usual, I solicit from all
readers, especially secondary school students, elegant solutions with possible
generalizations or extensions to all these problems. Readers submitting
solutions or comments should include their institution as well as clearly
identify the problems by giving their numbers and the year and page number of
the issue where they appear. The solutions need not be typewritten, but in
any case they should be easily legible.
I am continually grateful to receive problem sets from various national
olympiads around the world. It would be preferred if these were received in
English translation together with the official (or nonofficial) solutions
since this would make the editing job easier. In a number of cases in which I
did not have the translation and the solutions, I have had difficulty in
getting the problems down correctly. Two such examples occur in problems #6
and #12 following even though the problems were translated a second time by
another person. Consequently, if anyone comes up with the "correct"
formulation of these problems, I would be very appreciative to receive them.
show that among these three fractions two are equal to 1 and one is equal to
-1. (Grade 9, 10.)
8, Prove that for any natural number A", there is an integer n such that
connected by a line segment of the same color (no three segments intersect at
a point). Show that the red lines and the blue lines intersect in at least n
points, (Grade 10.)
• *
SOLUTIONS
First Round.
s a - 1
and thus we must have
a2m - a = a - 1 (mod 10). (1)
We now try a = 3, 7, 9 successively. If a = 3, (1) becomes 3 - 5 = 0
(mod 10) which is clearly impossible. If a = 7, (1) becomes 7 - 13 s 0
(mod 10). This is impossible since 7 = (50 - 1) 5 ±1 (mod 10). Finally,
if a = 9, (1) becomes 9 -7=0 (mod 10). This is also impossible since
2m
9 = (80 + if = 1 (mod 10).
Solution.
Here is a proof by induction, due to Golomb (Checkerboards and
polyominoes, A7»er. Math. Monthly 61 (1954) 675-682). Clearly the case n = 1
is valid. Assume the result is valid for n = k. Now consider the case
n = A' + 1, with the indicated cell C missing. Without loss of generality it
42
Solution.
Our proof is an indirect one. Assume that three collinear points A, Bf C
project into three noncollinear points A*, B', C'-. Let D' be a point on the
circle through A\ B*, C # and consider its preimage I?. It must be on the line
ABC or otherwise the two circles DAB and Z?BC would project into the same
circle A'B'C* which is contrary to the hypothesis. Thus the image of the line
AW is a circle. Now consider a circle tangent to line ABC say at A. The
projection of this configuration must be two circles tangent at A'. Consider
any line through A* intersecting the two circles in points P', Q'• The
preimages A, P, Q of A', P' and Q' are two points on line ABC and one off the
line. This gives a contradiction since the circle through A, P, Q must
project into a circle through A', P', Q'.
Solution.
Letting n = 2 , our proof is by induction on m. Clearly the result is
valid for m = 1. Assume the result is valid for n = 2 . Now consider the
case n = 2 . Since 2 -1 = 2 +2 4-(2 - 1), by the inductive
hypothesis we can always select three disjoint subsets, each of 2 numbers,
from 2 - 1 numbers such that the sum of each subset is divisible by 2 .
Letting the three sums bea-2 , 6-2 , c*2 , at least two of the numbers a, h,
c have the same parity. By selecting the two sets corresponding to these
numbers, we obtain 2 numbers whose sum is divisible by 2 . Consequently,
the result is valid for all positive integers m by induction.
A related result appears in H.B. Mann and J.E. Olson, Sums of sets in the
elementary Abeliam group of type (p,p), J. Combinatorial Theory 2 (1967)
275-284. The main result of this paper is as follows: Let /> be a prime, let
H be the additive group of residues modulo p and let G be the direct sum
- 44 -
Solution.
Since M is nonempty it contains at least one positive integer n, By
considering the sequence al, a 2f «** where a 0 = n and a . = [v5~7], j > 1, it
The solutions given here are the English translations (with small
changes) of the Hungarian solutions by L. Csirmaz*
Solution*, p
Case A. If there are two of the
points P, Q, R, say P and G? such that
segments AQ and BP intersect (at T, say), 4
then we have by the triangle inequality:
.4Q + PB = AT + TQ + PT + TB > AB + TO. (1)
Again by the triangle inequality,
AP ± AR > PR {2}
and
BR t BQ > RQf (3)
- 45 -
and by adding inequalities (1), (2), (3), we get the desired result.
Case B. The other possible
configuration is given in the figure.
Without loss of generality, P and Q
lie on the same side of AB extended.
B
It then follows that PQ extended must
cut segment AB. Now, without loss of
generality, Q will be contained in triangle BPR. Then,
PQ + QR < PB + BR,
AB < AQ + QB,
RP < AP + AR.
Adding, we again get the desired result.
Solution.
One can place the n rooks on the chessboard in nl different ways such
that no two rooks are attacking each other. Now we estimate the number of
these placings which do not satisfy the color condition of the problem.
Choose one of the nz/2 pairs of squares of the same' color. Put two rooks into
these squares. If these rooks are not attacking each other, this placing can
be completed in (n - 2)! different ways. Therefore there are at most
2
n (n - 2)!/2 wrong placings, and this number is strictly smaller than n\ for
n > 4.
Solution.
The remainder of the division n * d9 1 < d < 77 is one more than the
remainder in (n - 1) -f d, except if d is a divisor of n when it is d - 1 less.
Consequently,
r(k) « r(k - 1) = (A- - 2) - (sum of the proper divisors of A).
- 46 -
Thus w e wish to find infinitely many natural numbers k such that the sum of
the proper divisors of * is * - 2. This is the case if * = 2 W : the proper
ffl 2 ffi 1
divisors of 2 are 2, 2 f ... t 2 ~ f whose sum is 2® - 2.
T2$ . .. f T _-. Similarly, the sides BC and A£? are each divided into n equal
parts by the points Vlt U2t ... , 17 - (in order from B) and V l f V 2 f ... ,
Since B and C are linearly independent, we find that x = s/n and y = r/m.
Thus
P(r,s, = C{i • ££}B • « £ • a)c . (i)
Since P(r + l,s) - P(r,s) and P(r,s + 1) - P(r,s) are independent of r and s
respectively, each segment of the figure is divided into equal parts - m for
the ''horizontal" segments and n for the "vertical" segments (as shown in the
figure).
For the area \A(r,s)| of A(r,s) we have
2\A(r,s)\ = |{P(r + l,s) - P(r,s)} x {P(r,s + 1) - P(r,s)}|
Look now at any A(r9s) and its centro-symmetric A'(m - 1 - r,n - 1 - s).
From (2), (3) and (4) we have
\A(rfs)\ + |j'(» - 1 - r,n - 1 - s)\ = ~ . (5)
For n?f u odd it follows from this that the central small quadrilateral has
area A/ran. (The special case m = n = 3 was established using a long synthetic
proof by B. Greenberg, 7*af area problem, Math. Teacher 64 (1971), 79-80,)
If we take any small quadrilateral Q(r,s) and its centro-symmetric
quadrilateral Q(m - l - r s i 2 - l - » s ) w e see from (5) that
|Q(r,s)| + \Q(m - 1 - rtn - 1 - s) j = |<d(rfs) | + |-d'(r,s>|
-—+ — - — f
mn mn ~ mn
which proves part (b).
From (2), (3)f and (4) we have
\Q(r,s)\ = \A(rfs)\ * |4'<r,s)|
Solution.
Since only some of the triangles will contain P, a direct count is
difficult. So we consider all quadrilaterals formed by four vertices. For
each of these quadrilaterals P will lie in 0 or 2 of the four subtriangles of
- 49 -
Solution.
Clearly, the two subsets {1,3,5,...,2/2 + 1} and {n + 1,/? + 2,...,2n + 1}
are simple. Thus the maximum size of a simple set is > n + 1. We now show
that the maximum size is n + 1 by an indirect proof. Assume the maximum size
is at least n + 2 and let any n + 2 elements of this simple set be listed in
ascending order ax < a 2 < ... < a ~ = M. We will obtain a contradiction by
showing some a . and a . add to tf. Consider all the pairs from the initial set
Here are the answers to the 1985 Dutch Mathematical Olympiad, First Round
[1986: 2]. I am grateful to Andy Liu for providing them.
A. (1) 1585, (2) 1156, (3) 3971, (4) 320, (5) 1985, (6) 1/2.
B. (1) 168, (2) 5/2, (3) (3,7,42), (3,8,24), (3,9,18), (3,10,15),
(4,5,20), and (4,6,12), (4) a = 56, b = 334, c = 18704.
C. (1) a6/13, (2) 3456, (3) 286.
to Cecil Rousseau for passing on this information). Also, Greg Patruno notes
that Peter Yu's proposed problem [1985: 271, #5] appears as a special case in
J,P, Hoyt's Quickie Problem Q694, Math. Mag. (1984) 239, 242,
* * t
P R O B L E M S
%
1127. Proposed by D.S. Mitrinovic, University of Belgrade, Belgrade,
Yugoslavia.
(a) Let a, b, c and r be real numbers > 1. Prove or disprove
that
(log bef + (log,ca)r + (log abf > 3 • 2r.
a D c
(b) Find an analogous inequality for n numbers ax, a2, . . . , a rather
Triangles CBA and ADE are so placed that B and D lie inside ADE
• and CBA respectivel}-. DE and CB intersect at O. Angle CAB is equal to single
DAE. AD = AB and CO = OE. Prove that triangles ADE and CBA are congruent.
- 52 -
(a) Show that every positive whole number > 8*1 can be written
as the sum of three positive whole numbers in at least four ways (all twelve
numbers different) such that the sum of the squares of the three numbers in
any group is equal to the sum of the squares of the three numbers in each of
the other groups,
(b) Same as part (a), but with "three" replaced by "four" and "twelve"
by "sixteen".
t
(c) Is 84 minimal in (a) and/or (b)?
1130* Proposed by George Tsintsifasf Thessaliniki, Greece.
Show that
a3/2 + 53/2 + c3/2 i 3 7/ V /2
where a, b, c a r e t h e s i d e s of a t r i a n g l e and R i s t h e c i r c u m r a d i u s .
* * *
SOLUTIONS
No pio&lem l& evei permanently closed. T&e editor wLCC aCttHt-yn Ce pc&a&ed
to consider §on putllcatLon new notations on new insights on pant pnottcms.
of a. Prove that
n m n m
(a) I n a t . 1. . > 2 n fc * . 1Xj . ;
j=l 2=1 ^ j=l 2=1
n m n m
J
j=l i=l j=l 2=1 ^
f f =
par+1 V+2 ' •'• *W+DJ { V + l ' a wr+2 ?
"' ' a
ro(r+l)/ (1)
then
na = n 6 J ,
i ?7?r+i . mr+i
2= 1 2=1
and
- 55 -
r m r m
- J = l 1= i "O-D +i "J=1 J=1 rn(j-l) + l
is true by hypothesis, so (a) holds for n = r ,+ 1 . Thus assume (1) is not
true; then there exists 5 = 6 , , 1 < s <ro,so that
w+s
b £ la ,. , . . . , a , ,. , I
\ i?zr+l ' ' ffl(r+l)J
and t h e r e e x i s t s a = a , , , 1 < t < ro, so t h a t
mr+t ~
a
* { V + l ' ••• ' 5jD(r+l)/ '
since the two sets in (1) have the same size. It follows that b < a, since
the a.'s are in nondecreasing order. Also, a must equal
1 some b , . . V i . for
i ZH(J-1)+I
/ < r. Without loss of generality, take a = b . Let
©r
22?
B' = n- 6m(r-l)+i f
then
Jl—5 = B' < B = *-b
b mr b
and b < b mr
, so by the lemma we have
- mr
B' + B < £1—6 + £-6
- b b mr
mr
In other words, by interchanging b = a said b we do not decrease the right
side of (a). By continuing this kind of switching we arrive at (1), and
induction again takes over.
Part (b) is proved in a very similar fashion, using a similar switching
lemma: if x, y, u, v > 0, x + y < u + vf and u < y then (x + y)(u + v) >
(x + u)(y + v).
Part (a) (and also part (b)) implies the Arithmetic-Geometric Mean
Inequality. Given positive numbers .YX ,X 2 , . . . , x , put m = n and
h 1
i x . > n i n x- . ^
J J
j=i ' b=i J
For a companion result, see H.D. Ruderman, TVo new inequalities, Amer. Math.
Monthly 59 (1952), 29-32.
- 56 -
satisfy
12+1
(a
> i5= i T V i ;
73+1
a.
i
where the a. and b, are given positive real
i J
i=i
numbers«
1=1
given l a . = l , 0 < a . < l . This is a known result. In [1] it is shown that
n+1 1 - a
i , n+1
J a .- > n u;
i=l i
with equality if and only if a . = 1/n+l for all i. A simple proof follows
we o b t a i n
n+1
,1/n
n ( 1 - a,.)
i • n+1 a 1 a 2 . . . a n+1 72+1
i =l
= rr <? . ,
n+1
n 2=1 i=i J
equivalent to (1)•
- 57 -
Here is a related result due to Ky Fan [2]. Let alfa2$ ... fa be such
n 1 - x
that 2 a . = A with 0 < a . < 1/2 for all i. Since Cn is convex for
i=l J
0 < x < 1/2, it follows by Jensen's inequality that
n
1 -la.//}
fn11 - a . |> /}
1
i=l
I a
• I "
£n
I a /n
i
or
i? 1 - a. %/2
T—r i
1=1J ai . ?->
with equality if and only if a. = h/n for each i. As above, this is
obtain: if
then
1
n x. <
i=l 2
"
J"
n
with equality if and only if x. = ~- - 1 for each i.
1 A
l i =1 J
i =l
• (just replace x. by 1/x .). Also see M.S. Klamkin, Extensions of the
References:
[1] M.S. Klamkin and B.J. Newman, Extensions of the Weierstrass product
inequalities, Hath. Mag. 43 (1970)f 137-141.
[2] E.F. Beckenbach and R. Bellman* Inequalities, Springer-Verlag, New York,
1985, p,5.
corresponding
c x.i to be sufficiently small. If all a.
I
are greater than 1, the
k(y + x - z) = y * * x * . (y + x - z)
= y2 ~ z 2 ~ x 2 + 2.YZ
Z - X
z -x
= 2x.
This problem was taken from an extant 1893 Japanese sangaku, a wooden
tablet on which mathematical problems were written. During the 18 and 19
centuriesj sangaku had hung beneath roofs of shrines and temples throughout
Japan. Most problems recorded on them dealt with geometry, as was typical of
wazan, the Japanese mathematics of the time. Professor Fukagawa has
contributed quite a few of these problems, and they shall be appearing here
from time to time. They are likely unfamiliar and certainly beautiful, and
seem to be popular with readers. Professor Fukagawa has written a book (in
Japanese), Study of Sangaku, on this traditional Japanese mathematics; the
current problem is on page 138 of this book. He expresses an interest in
corresponding with people about geometry and its history. For readers wishing
to do so, here is his full address:
Professor Hidetosi Fukagawa
1-121 Higohara, Ogawa
Higasiuracho, Chitagun, Aichiken
470-21 Japan.
* * *
*=1 i=0
is a Fibonacci number of even subindex.
2U-1)
(-1)1 o 2 < * - ^ ) = o 2 ^ - D - G2<*-2> + ... + o" 2 ^- 1 1
i=Q
_2(Jt-l) 1 - (-G-2)2*-1
l +a
G2* + G~2k+2
rj
G~ + 1
from which
n 2U-1) " -,2k , ,,-2
V G +G
i i (.1)' o2'*-1-1' -
*=1 i=0
l - (f/V 1 - (Ci )
=2"
?TT 1 -G 1 -G
„2
£ _ [G2" - G~2n]
"G^TT
= J. [G2" - G"2n]
= F
2„ >
where we used
G4 - 1 (G - l ) 2 = 2G - 1 = 73"
following result.
Lemma. For each positive integer n,
2/? n, v
2 (-l)i G 2 1 " " 0 =F,2n+l '
1=0
Proof. Letting a = G and /J = (1 - Jo)/2 = -a , we have
- 62 -
2n+l 2n+l
a P
F = ~
2n+l ~
a - /3
G 2 « + l + (G-l)2n+1
rj
G +G
i=0
= F 4- F
2n 2/M-l
= F f
2n+2
completing our solution by mathematical induction.
The fact that lim {f(x) + af(bx)} = 0 means that for all € > 0 there is
x-+0
6 > 0 such that if x € A = (-5,6), then
&
\f(x) + af(bx)\ < e . (1)
Fix an arbitrary positive integer n. Let x be such that x and b x are
k
both in A . Then b x € A for A' = 1,2,...,ft and, from (1),
every n. Therefore we can conclude that (2) holds for every n. From (2) and
the inequality \f(b x) | < AT we get
\f(x)\ < e(l - la))""1 + \a\DH
for every n. Hence
|f(x)| < e(l - la])^1 + lim \a\nH = e(l - ( a p ^ 1 .
" - 1
|a|nK(y)| < e ^ 1 + tf
and h e n c e
|f(y)| < 4(|a| - l)"1 + \af"lt.
T h e r e f o r e , c h o o s i n g e < 1 a n d n = [e ] , we g e t
Then
f(x) + af(hx) = g(x) + ag(6x) + d ,
so
d = lim {f(x) + af(bx)} = lim {g(x) + ag(fcx)} + d .
x-*0 x-»0
Hence
lim {g(x) -i- ag(6x)} = 0 ,
x-0
and g is bounded in a deleted neighborhood of zero, so by the above proof
lim g(x) = 0. Thus lim f(x) exists and equals 1 .
a
x-*0 x-0
[Editor's note: We may as well complete the picture by noting that the
result is false for |a| = 1 and |b| * 0, 1. In this case, define
(-a) if x = b for some n € z ,
0 otherwise^
There has been the very good suggestion made that an issue of Crux
Mathematicorum be dedicated to Leo Sauve. I suggest the September issue of
this year. Accordingly, all readers who would like to propose problems
specifically for that issue, or have clerihews, limericks, rebuses, etc.
appropriate to the occasion, or would simply like to extend their best wishes,
are invited to submit same to the Editor. As many of your contributions as
possible will be published in September.
Bill Sands